To ensure the security and performance of traffic systems, managing vehicle flow and avoiding accidents at urban intersections, which are key locations, is essential. The smart intersection management system is presented in this paper. It uses the Internet of Things (IoT) and Artificial Intelligence (AI) to avoid collisions in real-time. This system used Recurrent Neural Networks (RNNs), which are great at predicting traffic patterns with sequential data. Data on vehicle speed, density, and movement is collected in real time by IoT sensors placed at key locations inside the intersection. An RNN model uses this information to make predictions about when traffic lights will go out and if they will remain on. The system can react proactively to possible dangers and adapt to changing traffic conditions due to the RNN-based proposed methodology. Intersection safety and traffic flow are improved by the system's ability to predict traffic patterns and identify potential collision situations. It uses the Collision Avoidance Challenge Dataset, which provides sensor data for predicting and preventing collisions in smart intersection systems using IoT and machine learning models. The system greatly decreases accident rates and increases traffic efficiency in evaluations conducted via field experiments and simulations. The RNN model achieved 96% accuracy in collision prediction, showing proficient real-time decision-making for smart intersection safety and markedly enhancing collision avoidance via IoT sensor data processing. © 2025 Elsevier B.V., All rights reserved.